Amc Webui
面向 Gemini 的 Local-First AI 工作流 WebUI，集成多模态聊天、Canvas、文件处理、实时搜索、代码执行与高级推理。
Last verified:
What is Amc Webui?
AMC WebUI is an All-in-one Model Console WebUI built with React 18 and TypeScript, designed primarily for Google Gemini native capabilities while also supporting OpenAI-compatible standard chat API. It follows a Local-First principle where chat data is stored in browser IndexedDB by default, ensuring privacy and smooth performance. The tool supports multimodal chat, Canvas, file processing, real-time search, code execution, and advanced reasoning features.
Key features include Deep Thinking (chain-of-thought visualization for Gemini 3.0/3.1/2.5 models with token budget and reasoning level settings), Live API for real-time bidirectional audio/video streaming with screen sharing, Live Artifacts that auto-render code blocks as interactive HTML with Mermaid and Graphviz support, and advanced file processing supporting ZIP/folders, images, PDF, video, and audio with configurable upload strategies. The tool includes a local Python sandbox based on Pyodide (WASM) with pre-loaded numpy, pandas, matplotlib, and on-demand package installation.
AMC WebUI is ideal for developers, AI researchers, data scientists, and power users who work extensively with Google Gemini models and need advanced features like deep reasoning, code execution, file analysis, and real-time audio interaction. It supports enterprise-grade API management with multi-key rotation, dual API modes (Gemini native and OpenAI-compatible), and can be deployed via standard Vite development, Docker Compose, or Cloudflare Pages with independent API service.
The tool offers productivity utilities including Deep Search aggregating Google Search with automatic task planning, URL context extraction, TTS with 30 voice options, voice transcription, Imagen 4.0 image generation with Fast/Standard/Ultra tiers, and comprehensive data management with chat import/export, session grouping, and full-text search. It includes PWA support for desktop/mobile installation, multi-tab synchronization via BroadcastChannel, customizable shortcuts, and enterprise security filters for harassment, hate speech, pornographic content, dangerous content, and civic integrity.
Amc Webui pricing
Pricing model: Freemium
AMC WebUI is completely free and open-source under the MIT license. There is no paid tier or subscription. The tool itself costs nothing - users only pay for what they spend on LLM tokens through their own API keys (Gemini API or OpenAI-compatible services). The application supports BYOK (Bring Your Own Key) mode where users provide their own API keys. Server-managed API key deployment is optional for self-hosted environments. Docker deployment defaults to BYOK mode where users enter their Gemini API Key in settings after launch. No hosting fees are required as the tool can be self-hosted via Docker Compose, Cloudflare Pages (free tier), or static deployment.
Amc Webui pros
- Local-First architecture with IndexedDB storage ensures privacy and offline capability
- Native Gemini support with Thinking/chain-of-thought visualization for deep reasoning
- Live API enables real-time bidirectional audio/video streaming with screen sharing
- Local Python sandbox (Pyodide WASM) with numpy, pandas, matplotlib pre-loaded
- Live Artifacts auto-render code blocks as interactive HTML with Mermaid/Graphviz
- Advanced file processing supports ZIP, folders, images, PDF, video, audio
- Deep Search aggregates Google Search with automatic task planning and citations
- Dual API modes: Gemini native and OpenAI-compatible with separate key configuration
- Multi-key rotation support for both Gemini and OpenAI-compatible APIs
- Imagen 4.0 image generation with Fast/Standard/Ultra tiers and aspect ratio control
- TTS with 30 voice options and multi-model voice transcription support
- PWA support with installable desktop/mobile app and offline shell
- Multi-tab synchronization via BroadcastChannel with Web Locks API protection
- Comprehensive data management with chat import/export, session grouping, full-text search
- Customizable keyboard shortcuts for all major operations
- 5 security filter categories with configurable blocking levels
- Bilingual interface (Chinese/English) with system theme following
- Developer log panel with API call monitoring and token usage statistics
- Docker Compose deployment with web+api dual container setup
- URL context extraction automatically fetches webpage content for conversation context
Amc Webui cons
- Requires Node.js 26 specifically (engine-strict will fail on 25 or 27+)
- Live API requires browser-local API key, limiting true server-managed deployment
- OpenAI-compatible mode does not use Gemini native toolchain
- Security boundary designed for trusted/self-hosted environments, not public multi-user
- Pyodide large assets excluded from PWA pre-cache require network on first Python execution
- Multi-key rotation requires manual configuration in settings interface
- MCP stdio and internal HTTP access disabled by default for security
- No built-in user authentication for public deployment without additional setup
- OpenAI-compatible mode model list must be manually maintained
- API proxy configuration requires understanding of multiple runtime environment variables
- No native mobile app (relies on PWA for mobile experience)
- History without precise pricing metadata shows dash instead of cost
- Cross-origin deployment requires manual ALLOWED_ORIGINS CORS configuration
- Third-party AIStudioToAPI integration requires self-evaluation of security risks
- Image generation pricing metadata may be incomplete for some request types
Frequently asked questions about Amc Webui
What is AMC WebUI and what does it do?
AMC WebUI is an All-in-one Model Console WebUI built with React 18 and TypeScript, primarily designed for Google Gemini native capabilities while also supporting OpenAI-compatible standard chat API. It integrates multimodal chat, Canvas, file processing, real-time search, code execution, and advanced reasoning. The tool follows Local-First principles with chat data stored in browser IndexedDB, supporting deployment via Vite development, Docker Compose (web+api), or Cloudflare Pages with independent API service.
Is AMC WebUI free to use?
Yes, AMC WebUI is completely free and open-source under the MIT license. There are no paid plans or subscriptions. The only cost is whatever users spend on LLM tokens through their own API keys (Gemini API or OpenAI-compatible services). The tool supports BYOK (Bring Your Own Key) mode where users provide their own API keys, and Docker deployment defaults to this free mode.
What AI models does AMC WebUI support?
AMC WebUI supports Google Gemini 3.x series (gemini-3-flash-preview, gemini-3.1-flash-live-preview, gemini-3.1-flash-lite, gemini-3.1-pro-preview), Gemini Robotics, Gemma 4 models, Imagen 4.0 for image generation, and Gemini TTS models with 30 voices. The OpenAI-compatible mode supports any model available through OpenAI-compatible chat/completions endpoints including OpenAI official, Gemini OpenAI-compatible endpoint, or other compatible services.
How do I deploy AMC WebUI?
There are three main deployment methods: (1) Standard development mode using Vite locally with npm run dev (requires Node.js 26), visiting http://localhost:5175; (2) Docker Compose mode with web+api dual container deployment, running npm run build:docker followed by docker compose up -d --build, accessing http://localhost:8080; (3) Cloudflare Pages for static frontend plus independent Node API service deployment, with RUNTIME_API_PROXY_URL configured to point to the backend.
What is the Deep Thinking feature?
Deep Thinking supports chain-of-thought visualization for Gemini 3.0/3.1/2.5 series models. Users can set Token Budget or Reasoning Level (Minimal/Low/Medium/High) to control the depth of reasoning. The feature allows real-time viewing of the AI's logical calculation process, making the model's reasoning steps visible during generation.
How does the Local Python sandbox work?
The Local Python sandbox is based on Pyodide (WASM) running in a Web Worker within the browser. It comes pre-loaded with numpy, pandas, and matplotlib, supports automatic dependency detection and on-demand installation of packages like scipy and scikit-learn, enables file mounting and generated file downloads, and automatically captures matplotlib chart outputs. All code execution happens locally in the browser without sending code to external servers.
What is the Live API feature?
Live API provides bidirectional real-time streaming interaction with voice call support. It includes screen sharing and visual recognition, audio visualization using AudioWorklet API, and uses browser-local API key to directly connect to official Live WebSocket (not routing through AMC backend). This enables low-latency audio/video streaming for conversational AI experiences.
Can I use AMC WebUI with OpenAI models?
Yes, AMC WebUI supports OpenAI-compatible standard chat API mode. This mode uses independent API Key, Base URL, and model list separate from Gemini native configuration. Requests are sent to POST {Base URL}/chat/completions, supporting both normal and streaming responses with system prompts, temperature, top_p parameters. Compatible with OpenAI official interface, Gemini OpenAI-compatible endpoint, or any chat/completions-compatible service. Base URL examples include https://api.openai.com/v1 for OpenAI official.
How do I handle file uploads in AMC WebUI?
AMC WebUI supports advanced file processing with browser-side audio preprocessing/compression to reduce token and bandwidth consumption. You can drag ZIP/folders for automatic code repository structure parsing. Supported file types include images, PDF, video, audio, and text. Users can configure whether each file type uses Gemini Files API or direct Base64 upload. Input detail level is adjustable (Unspecified/Low/Medium/High, with Ultra High available for single images).
What security features does AMC WebUI have?
AMC WebUI includes 5 security filter categories: harassment, hate speech, pornographic content, dangerous content, and civic integrity. Each category can be independently configured with blocking levels (Off/Don't Block/Block Few/Block Some/Block Most). The tool follows Local-First principles with data stored in browser IndexedDB for privacy. For public deployment, additional authentication, quota/rate limiting, abuse protection, audit, and tenant isolation are recommended beyond the default trusted self-hosted configuration.